Researcher, Developer · My major contribution was proposing an Ant Colony Optimization based feature scoring algorithm for selecting the most important… · More features from the dataset. I also tuned and tested the classifiers and selected Support Vector Machine as the primary classifier and Neural Networks to be the secondary classifier. I was also responsible for all the programming that was needed for the project.
The focus of this project was to be able to efficiently detect and prevent network intrusions and also classify the type of the attack for report generation. This required the implementation of two separate classifiers and a proposed feature selection algorithm.
Finding candidates interested in a job by mining emails.
Data Scientist, Developer (Intern), Python Scikitlearn, Python/Flask · The purpose of the project was to understand which candidates were interested in… · More a job offer by going through e-mail replies in the database. This would enable TA's to quickly identify the candidates and prioritize their engagement with candidates without having to go through their emails one by one.
I was given the task of building the entire system. This required me to put up data extraction and labeling system which was used to generate the train and hold out set for the classifier. I wrote an API that could preprocess the data, feed the data to the classifier, tune the parameters of the classifier and give predictions for a given e-mail body when passed as a JSON object. I used a bag of words approach to generate features for the classifier. The final API was deployed as a python Flask application.
Assigning multiple suitable verticals to candidate profiles
Data Scientist (Intern), Python Scikitlearn · The purpose of the project was to identify the most suitable verticals (such as Software Engineer,… · More Analyst, Testing Engineer) for a given candidate profile.
My main job was to come up with a machine learning model that can generate multiple suitable verticals for a single candidate profile. Most of my work was purely experimental. I developed a K Nearest Neighbors based approach to solve this problem
Developer, Python, Numpy · RNTN - Python-numpy only implementation of Recursive Tensor Neural Network
I derived the derivatives of the… · More parameters of an RNTN by hand and coded the module from scratch. The RNTN is optimized by Adagrad (also coded by me) and uses "Backpropagation Through Structure" (BPTS) as the learning method.
What I Do
Aspiring data scientist. I have moderate experience with Machine Learning solutions and looking forward to broaden my skills.